Sie sind auf Seite 1von 13

Super-Cavitating Profiles for Ultra High Speed Hydrofoils:

a Hybrid CFD Design Approach


Stefano Brizzolara1 and Alessandro Federici2
M a r i n e C F D G r o u p , Genova - Italy
1
University of Genova, Italy, Department of Naval Architecture, Head of Marine CFD Group
2
University Pole for Yacht Design of La Spezia, Italy

ABSTRACT:
The paper presents the main results obtained from a systematic assessment about the current possibility to
simulate hydrodynamic characteristics of super-cavitating hydrofoils with state of the art CFD methods. First
a systematic validation of the finite volume RANSE solver with volume of fluid method for the multiphase
flow and a simple Rayleigh-Plesset model for bubble dynamics. Sensitivity of the solver to various
parameters which affect the cavitation model is verified as well as the integration time-step, the number of
inner iterations, which influence the unsteady calculations. The final best configurations of the CFD model
are used for its validation against experimental results on a reference two terms super-cavitating profile at a
typical design angle of attack and for the complete range of cavitation indexes. After this preliminary study,
the paper continues with the verification of the performance in case of different super-cavitating hydrofoils
with two/three and five terms face shapes, designed using a classical asymptotic theory for the face and the
back shapes chosen on the basis of cavity shapes predicted RANSE simulations. The RANSE method proves
to be effective to design the hydrofoil back side for finite cavitation numbers and to assess its performance of
for the next 3D hydrofoil design.

NOMENCLATURE

BOA

vessel breadth over all [m]

p0
pV
V
VoF
CL ,D
c

hydrostatic (ambient) pressure at foil depth [Pa]


vapour tension at given water temperature [Pa]
flow speed [m/s]
volume of fluid
Lift or Drag Coefficient 2(L , D)/(cV2)
hydrofoil chord length [m]

V,0 2(p0-pv)/(V2), inflow speed cavitation number

water mass density [kg m-3]

2 INTRODUCTION
Results presented in the paper are those obtained in
the preliminary phase of more ample research
project dedicated to the concept design and
hydrodynamic optimization of a super-high speed
autonomous surface vehicle developed by the
Marine CFD Group of the University of Genoa in a
IX HSMV Naples 25 - 27 May 2011

research program sponsored by the Office of Naval


Research Global. As from general specifications, the
unmanned vehicle is meant to reach a top speed of
120 knots in sea state two, but is also requested to
loiter at slow speed for days. To this scope, an
unconventional SWATH hull has been designed for
loitering mode and optimized for minimum
resistance at take-off speed, on the basis of the
previous experience in the design of unconventional
SWATH vessels (S. Brizzolara, 2004) (Brizzolara &
Bruzzone, 2007) also in case of unmanned vehicles
(Brizzolara et al. 2011). For the operations at super
high speed, the unconventional displacement hull
can be converted into a hybrid hydrofoil-wig craft
by means of two pairs of retractable super-cavitating
hydrofoils. The initial concept design of the vessel is
presented in the two picture of Figure 1, which
shows 3D renderings of the two operating modes:
hullborne and foilborne.
The design of the surface piercing hydrofoils turns
to be one of the most challenging design topics of
the project connected with the ultra-high speed
1

requirement. It is well known, in fact, that cavitation


becomes almost unavoidable at speed greater than
50/60 knots, above which super-cavitating hydrofoil
shapes have to be used.
The major hydrofoil studies and scientific
advancements were done between the 60s and the
80s and indeed few example of hydrofoils vessels
with top speed higher than 60 knots can be found
from those years, as for instance super-cavitating
submerged hydrofoils TAP-1 or TAP-2 designed by
Boeing (Baker E., 1975).
The design methods for super-cavitating hydrofoils
are still nowadays primarily based on the linearised
asymptotic theories developed in the mid of last
century by Tulin and Burkart (1955), generalized to
the design of 3D hyofoils near the free surface by
Virgil Johson (1957). This theory is valid only in the
limit of zero cavitation number and is able to give
the shape of only the face of the profile (since the
back is supposed to lie inside the supercavity, by
definition). Actual profiles for the previously
mentioned super-high speed vehicle have to be
optimized for a finite cavitation number (0=0.05 at
120 knots) and their back face need to lie inside the
cavity while ensuring a sufficient thickness for
strength issues.

In this respect, the authors have decided to explore a


mixed design approach: the linearised theory to
design the face and a CFD approach to verify the
hydrodynamic characteristics and design the back
side with a trial and error procedure. Hence the topic
of RANSE solvers of multiphase cavitating flow.
The solution of the flow field around supercavitating hydrofoils is rather complex because of
the critical and unstable nature of cavitation. Modern
volume finite volume unsteady RANS solvers
promise to be a valid tool to study this complex
phenomenon, but the theoretical models need to be
set up correctly for what regards the parameters that
control the physical models and numerical
approximation schemes.
After a concise description of the theoretical and
physical models used by the numerical method, the
paper will continue with the description of its
validation on a set of experimental results for a 2D
super-cavitating foil in section 4 and finally will
illustrate the design method followed to design the
basis 2D section of the surface piercing supercavitating hydrofoils of the autonomous vessel.

3 MULTIPHASE VISCOUS FLOW SOLVER


3.1 RANSE solver with multiphase VoF model
A state of art of RANSE solver with VoF method for
representing the free surface has been selected. The
software suite (CD-Adapco, 2009) has the capability
of solving model or full scale turbulent flows around
a body in non-stationary conditions, with the VoF
method for predicting the free surface around it, with
a simplified adiabatic cavitation model, described in
the next paragraph.
The solver is applied to the following group of
equations which express the mass and momentum
balance with an Eulerian approach and Reynolds
time-Average approach with the needed boundary
conditions valid for this type of CFD simulation.
The RANS equations can be expressed, in our case,
for an incompressible flow as shown in Eq. 2:

V 0

V P V TRe S M

Figure 1 Initial concept (patent pending) of the hybrid


Hydrofoil-SWATH-WIG in foilborne (up) and hullborne
(bottom ) operating modes

IX HSMV Naples 25 - 27 May 2011

(1)

where V is the Reynolds averaged flow velocity


vector, P is the average pressure field, is the
dynamic viscosity, TRe is the tensor of Reynolds
stresses and SM is the vector of momentum sources.
The component of TRe is computed in agreement
with the standard k- turbulence model selected for
this application, in agreement with the Boussinesq
hypothesis:
2

Vi

V j 2
k ij
3

x
j
i

2
2 t Dij k ij
3

ijRe t

3.2 Cavitation Model

(2)

Where t is the turbulent viscosity, k is the turbulent


kinetic energy. The realizable k- turbulence model
was selected to close the hydrodynamic problem. To
save cells close to the hull surface, an analytical wall
function has been adopted to the velocity vector and
all other scalar quantities are extrapolated from the
known quantities on the wall boundary surface.
As regards the wall function, on the cell closest to
the profile, a two layer model approach has been
applied. The two layer wall function model is a
model that imposes a first thin laminar layer near the
wall and a second logarithmic layer over the first;
this model assumes that the centroid of the first cell
near the wall lies within the logarithmic region of
the boundary layer. The wall treatment is optimized
to approximate boundary layer effects with a
dimensionless wall distance y+<100, but a much
smaller value has been used in this study in order to
accurately solve the cavity thickness growth along
the wall.
The RANS solver is based on a Finite-Volume
method to discretize the physical domain. The
equation for an incompressible multiphase fluid is
used in the simulation, with one more transport
equation for the VoF, shown in Eq. 3, which
represents the fraction of water inside each cell:

VoF
VoF U 0
t

(3)

This equation ensures to find the correct shape of the


free surface between water and air, this method is
powerful for solving problems when the wave
breaking may occurred, or when the air effects are
important on the free surface shape.
To solve the time-marching equations, an implicit
unsteady solver is used.
All hydrodynamic unknown quantities in the field
are solved at each time step using an iterative
method. In particular, many inner iterative steps are
needed to reach a good prediction of the unsteady
cavity evolution at each time step. The software uses
a SIMPLE method to conjugate pressure field and
velocity field, and a AMG (Algebraic Multi-Grid)
solver to accelerate the convergence of the solution.
A more complete description can be found in the
technical reference manual (CD-Adapco, 2009).

IX HSMV Naples 25 - 27 May 2011

The Cavitation model is based on the flow mixture


concept and the Rayleigh-Plesset simple model to
simulate bubbles dynamics. With this simplified
model one can define the cavitation inception and
bubble dynamics by setting the size and number of
micro-bubble seeds (cavitation nuclei), and their
stochastic spectral distribution in the considered.
These are approximately represented in the RANSE
solver through an average seed radius R0 and an
average seed density n0. The latter is a constant
strongly dependent on water quality and it is defined
as the number of cavitation seeds (nuclei) per unit
volume of liquid (Sauer 2000).
To simplify the problem, the vapor bubbles in a
control volume are represented in an average sense
by an homogeneous distribution of seeds with the
same mean radius. This assumption allows to
describe the bubble distribution by a single scalar
field, the vapor volume fraction V, which
influences directly the VoF variable solved in (3).
Assuming that only one liquid phase and the
corresponding liquid-vapor phase can occupy the
control volume where cavitation takes place:
V

n 4 R 3
VolV
034 3
Vol l VolV 1 n0 3 R

(4)

where Volv is the volume occupied by the vapor, Voll


the volume occupied by the liquid and Nbub the
number of vapor bubbles in the control volume.
This vapor volume fraction changes, as all phases,
due to convective transport but also to bubble
growing and collapsing.
The modeling of bubble grow rate is based on a
Lagrangian observation of a cloud of bubbles, using
the Rayleight-Plasset equation to describe the
average bubble radius time evolution R(t):
p p0 2
dR
d 2 R 3 dR
v

4 l
2
dt
2 dt
l
l R
l R dt
2

(5)

where pv is the saturation pressure corresponding to


the temperature at the bubble surface and l , l are
the water density and viscosity respectively, and is
the surface tension coefficient.
The model is simplified and the final equation for
the bubble growth, obtained neglecting the inertia,
viscous and surface tension terms in (5) is given by
the following expression:
2 pv p0
dR

dt Ray 3 l
2

(6)

4 ACCURACY OF THE RANSE MODEL


Different numerical models have been generated to
be used with the RANSE solver described in the
previous section, by systematically varying
geometrical mesh parameters as well as cavitation
model constants, to be validated against the
experimental results on a super-cavitating hydrofoil
of Waid & Lindberg (1957).
The systematic validation study has been performed
for different angles of attack of the profile (6, 4 and
2 degrees), for each of them exploring a quite wide
cavitation index range, by keeping constant the flow
speed and opportunely changing the ambient
pressure. Only the results of the 6 degree are
reported in this article, as most representative of a
typical design conditions.
The general approach, derived from experience in
order to cope with particular instability issues of the
unsteady cavitation model, is first to converge on the
steady non-cavitating viscous flow field around the
airfoil and then to gradually increase the vapor
pressure to its correct value from an initial (virtual)
very negative one. This procedure is necessary to
avoid the cavitation model to be applied on the
initial unrealistic negative pressures field which rise
for the sudden acceleration of the flow around the
foil in the first instants, during the solution of the
virtual transient to converge on the steady flow field;
these unrealistic negative pressure values, which
gradually disappear until convergence on the steady
non-cavitating flow field around the foil, would
cause the unrealistic formation of large vapor
bubbles around the foil and would lead the solution
to divergence.
4.1 Physical model set up
The set-up of the parameters which rule the physical
model has been also systematically investigated.
First, as perhaps obvious, for the inherent unsteady
nature of cavitation an unsteady solution must be
searched for in order to find a feasible solution,
although the final global parameters compared with
the experimental tests are the mean forces
components. Also the integration time-step of the
solver was studied: a constant value of 110-3 s was
selected as the minimum for stability and accurate
results. In fact, higher time-step values brought to
inaccurate results, while smaller values made the
convergence property of the solution very weak
being too sensible to the turbulent fluctuation of the
flow around the foil.
As regards parameters of the cavitation model, in a
particular case for 0=0.48 (i.e. the region where the
highest differences between experimental and
IX HSMV Naples 25 - 27 May 2011

numerical results are registered) it was verified the


effect of a variation in the seed diameter of the
cavitation nuclei (chapter 3.2) from 1e-6 to 1e-4
meter; this test is carried out with the mesh type Q2
(see chapter 4.2). The results are quite interesting: as
the seed diameter increases, the lift coefficient CL
decreases coming closer to the experimental value,
on the other hand CD instead diverges. A good
compromise seems to lay around 510-5 m, in fact the
accuracy on CL increases of about 4% while that on
on CD is worse of about 2%. Also the ratio L/D at
this point is quite in line with the tunnel tests results.
Not being convinced about the obtained trends and
being uncertain of the tunnel tests at this particular
cavitation index and about the effect of this
parameter for all the other 0, the more typical seed
diameter value of 510-5 m was assumed, eventually
privileging a smaller error on drag than that on lift.
On the other hand, specific measurement of the
cavitation nuclei size statistical distribution were not
made in the reference tests.

Graph 1 Influence of varying bubble seed diameter on the


Lift and Drag Coefficient =6 degree, 0=0.48)

Other sensibility tests varying the seed number Nbub


were performed, but this parameter does not
appreciably influence the results, so the standard
value 1012 nuclei per cubic meter of water was
assumed for all the simulations.
4.2 Mesh sensitivity First Systematic Analysis
As a result of this systematic analysis a domain
spanning of 3 chord lengths upstream of the foil
leading edge, 15 chord lengths aft of the trailing
edge and 5 chord lengths from the face and back
sides of the foil itself has been chosen, as illustrated
in Figure 2. This with the scope of eliminating the
influence of the boundaries (especially the inlet and
top and bottom slipping walls) on the solution. Since
results reported from the tests in the water tunnel on
a small model (3x3 inches) were not corrected
blockage and upper/lower wall effects, some

uncertainty remains when comparing the forces


which needs to be discussed case by case.
To find the correct mesh refinement, two different
sensitivity analyses were performed: the first with a
closer inlet, about 1 chord lengths upstream of the
foil (usually sufficient for standard profiles), the
second with 3 chord lengths; two different
refinements have been applied to both mesh sizes.

The thickness of the prism layer plays an important


role: this must be greater than the cavitation bubble
height in the condition that has to be solved. Figure
5 shows the difference between two different prism
layer sizes at the same condition, the difference in
results are quite important: not only we have a
reduction of the cavity length but also a different
unsteady behavior of the bubble (and induced
forces) with time. This problem is rather complex
and is further detailed in paragraph 4.3.

5 Chords
3

15 Chords

5 Chords
Figure 2 Fluid domain, final sizes (mesh Q2+I)

Figure 3 gives the idea about how large must be the


refinement around the foil in order to correctly
capture the shape of the cavity. In fact, a thick prism
layer of very thin cells is needed over the suction
side of the hydrofoil.
Figure 5 Effect of two different prism layer thicknesses on
the predicted cavity shape (same condition)

Figure 3 - Mesh Q2+I around the foil

Figure 4 Close up on prism layer mesh near the leading edge


of the foil and results of the VoF of the liquid phase.

The close up of Figure 4 renders the quality of the


solution of the vapor volume fraction in terms of
free surface tracking capability along the prismatic
cells generated on the suction side of the foil.

IX HSMV Naples 25 - 27 May 2011

The development of the cavity as a function of the


cavitation index is given in the pictures of Figure 21
in the Appendix. The effect of the mesh refinement
is also important when the cavity overpass the foil
length (super-cavitating conditions): a proper
refinement behind the trailing edge is also needed to
catch the correct cavity free surface and the unsteady
turbulent flow in the cavity closure region. Even at
the highest cavitation indexes (=1.71.3), as soon
as the partial cavity is forming and growing at the
leading edge, it gradually (and significantly) affects
the hydrodynamic predicted forces, since it alters
considerably the distribution of pressure and flow
field in the surrounding areas (see Figure 19 and 20).
In fact, both lift and drag force components are
influenced by the partial cavity development and as
from Graph 3 and Graph 4. In these graphs four
different numerical curves are plot corresponding to
for four different meshes considered in the
sensitivity analysis. The experimental values are
taken from the tests of Waid & Lindberg (1957) on a
3x3 inches wing spanning the full width of the
cavitation tunnel test section and without any

correction for blockage or top and bottom wall


reflection effects.
The first mesh type, named Q1 corresponds to the
lighter mesh with the lowest number of prism layer
cells from the wall and a refinement region of one
chord behind foil trailing edge. The second mesh
type, Q2, is more refined but has the same
extensions of mesh Q1.
As expected, the main differences between the
global forces predicted in both cases are around 0 =
0.5, where the cavity is partially interesting the
profile and particularly unsteady. Indeed, it is
troublesome to solve the flow in these conditions:
due to the inherent instability of the phenomenon, in
fact, the bubble grows up, bursts and regenerates
itself continuously and this high frequency behavior
is not simple to follow numerically. More details
will be given in the next paragraph.
At lowest cavitation indexes, i.e. in supercavitaing
conditions (0<0.45 at the considered angle of attack
of 6 degree), the predicted forces are well following
the trends of the experimental measurements, still
showing considerable errors on experimental CL
(about +15%) and on CD (around 22%).

the measured forces should be constant, while in fact


they have a fluctuation both on lift and drag
components. Perhaps the order of magnitude of the
measurement precision is in this range of these
fluctuations. Focusing the attention to the fully
cavitating condition (0 < 0.6), the best results are
achieved with the Q2+I mesh type, both for CL and
CD. The average error in this regime being an
overestimation of about 8% on CL and 12% on CD.
As anticipated in the previous paragraph, the
problematic zone is between 0 =0.5 and 0 =1.0; A
deeper analysis of the problem revealed that in these
cavitating conditions the cavity forming on the foil
suction side is not stable, but it is growing and
collapsing at a certain frequency.

4.3 Mesh Sensitivity - Second Systematic Analysis


After assessing the influence of the mesh refinement
from first sensitivity analysis (in particular on the
pressure field around the foil) a larger domain
extension was tested since some influence of the
boundary conditions on the local solution around the
foil was noted. The new domain size corresponds to
the extensions quoted in Figure 2.
So also in this second sensitivity analysis the two
different mesh resolutions were investigated but
with a larger domain size: the new codes for these
two meshes are Q1+I (larger size but lower
refinement) and Q2+I (larger size and higher
refinement). The forces obtained in these cases are
represented in Graph 3 and Graph 4, with orange
and red lines, respectively.
As clear the difference in lift and drag coefficient in
non cavitating conditions (0=2.95) has been solved.
The shift of the inlet from one to three chord lengths
upstream of the profile is the main responsible of the
good correlation with the experimental lift force,
regardless of the mesh refinement. For the drag,
instead, also the mesh refinement influences the
predicted force and strangely the coarser mesh is
closer to the experimental results in non-cavitating
or partially cavitating conditions than the finest one.
It is to be noted that also experimental measurements show some unexpected result: in fact for
01.5 when there is still no cavitation on the foil,
IX HSMV Naples 25 - 27 May 2011

Graph 2 CL Time history for two different mesh refinements

Graph 2 presents the time histories of the lift


coefficient prediced with two different mesh
refinements at V = 0.744. The minimum cell size
condition the frequency with which the lift force
signal oscillates (and so the cavity on the back).
With a low refinement, after an initial period of time
in wich the lift is growing similarly to the more
refined mesh, we note a certain quiescent period in
which the force remain almost constant (3.2s to
5.0s). This phenomenon is differently predicted with
the higher resolution mesh, for which the cavity is
always contained within the prism layer height
during its growth. The finer resolution mesh is able
to capture a periodical growth and collapse of the
cavity (so the lift) on the hydrofoil suction side,
being more adherent to reality (Brennen, 1995). This
different behavior with time obviously influence
also the mean value of the force when averaged over
a period of about 5 seconds, which is the value
compared with the experiments. So the apparently
better correlation of CL with the Q1 mesh is due to
this unphysically predicted lift time history and not
to a real improvement in the model accuracy.
The high frequency oscillations of the partial
cavitation stops at lower cavitation numbers, as soon
as the cavitaty interests the complete length of the
profile.
6

CL

0.95

0.85

0.75

0.65
CFD Q1
0.55

CFD Q2
CFD Q1+I

0.45

CFD Q2+I
Tunnel

0.35

0.25
0.0

0.5

1.0

1.5

2.0

2.5

3.0

Graph 3 - Lift Coefficient sensitivity to different mesh types and resolution

CD

0.07

0.06

0.05

Tunnel
CFD Q1
CFD Q2

0.04
CFD Q1+I
CFD Q2+I

0.03
0.0

0.5

1.0

1.5

2.0

2.5

3.0

Graph 4 - Drag Coefficient sensitivity to different mesh types and resolution

IX HSMV Naples 25 - 27 May 2011

A similar behavior is predicted for the drag


coefficient (Graph 5), which can be subject to the
same error in averaging the unsteady signal. The
peaks and through of lift are synchronized with
those of drag, so the cavity affects the two force
components in the same way.

to standard circular arc or Tulin-Burkart profiles, but


at the cost of more complex line shapes.
As presented in Graph 6, in fact, the three and five
terms shapes have a higher maximum camber that is
moved further aft with respect to the corresponding
two terms shape and even an initial negative camber
more pronounced in the five terms shape.
0.07

0.06

Tulin-Burk

0.05

J3terms

0.04

J5terms

y/c

0.03
0.02

0.01
0
-0.01 0

0.2

0.4

0.6

0.8

-0.02
-0.03

Graph 5 - CD time history for two different mesh refinements

Additional Figures of the results obtained from the


systematic CFD calculations on the mesh Q2+I are
reported in the Appendix, where Figure 19 presents
streamlines and pressure field around the foil and
Figure 20 the flow velocity magnitude.
From these figures it clearly distinguishable the
separated zone aft of the brunt trailing edge which
are interested by recirculating flow in wetted as well
as cavitating conditions. Also it is possible to note
the constant pressure inside the vapor cavity and
behind the blunt trailing edge of the hydrofoils.

DESIGN & OPTIMIZATION OF A NEW FOIL

After this initial investigation devoted to the tune


and set-up of the CFD model, the study continued
with the design and optimization of the 2D hydrofoil
to be used as a base section for the 3D
supercavitating hydrofoils which sustain the vessel
at high speed.
The approach followed was a hybrid approach: first
the face of the supercavitating hydrofoil was
designed with traditional linear asymptotic theories
(V. Johnson, 1957) with the desired CL value. In
fact, in fully cavitating condition, only the face carry
out the lift, because the back and the trailing edge
are normally covered by the cavity vapor bubble.
The theory developed by Johnson (1957) is based on
the conformal mapping method of Wu, as linearised
developed by Tulin and Burkart (1955), generalizing
the series expansion which defines the vorticity
distribution along the chord of the profile (in the
imaginary plane) to higher order and so opening the
possibility to obtain higher efficiencies with respect

IX HSMV Naples 25 - 27 May 2011

x/c

Graph 6 Face shapes obtained for different super-cavitating


profiles shpes at the given design CL=0.25.

As a result of the conformal mapping techniques


extended with higher order terms the maximum
theoretical efficiency which can be obtained in ideal
conditions (zero angle of attack) for the two terms
profile (also called Tulin-Burkart) is:
L 25

D 4 2C L

(7)

while the Johnson three terms profiles have 1.44


times higher efficiency than that (7) of the TulinBurkart shapes and the five terms profile can reach
1.78 higher efficiency at the same lift coefficient.
The cited design method developed by Johnson
extends even further these asymptotic methods valid
for 2D sections which are valid for infinite water
depth, by considering also free surface effects, and
the three dimensional effects caused by the finite
aspect ratio of a 3D hydrofoil and that due to the the
cross flow.
Based on the cited theory, three different 2D face
shapes have been designed with a 2D-CL=0.42,
corresponding to a 3D-CL=0.25 for the surface
piercing foil, according the approximate corrections.
The design angle of attack of 5 degree (higher thean
the ideal one) was selected in order to have a
sufficient cavity thickness at the design operating
point. After having defined the face shape, also the
shape of the back have been designed with the aid of
RANSE simulations by a trial and error method,
better described herein after. Moreover RANSE
simulations allow also to verify the order of
approximation of the asymptotic theory in
evaluating the theoretical lift coefficient for 0=0,
8

instead of the actual design cavitation number


0=0.05 (at the top speed of 120 knots). At finite
cavitation indexes, the real cavity thickness is
important and the shape of the back must be
designed in order to obtain the maximum profile
thickness at each chord station (to increase the
section area modulus, for strength reasons) while
ensuring that the back line is still laying inside the
vapor cavity. From the comparison of the designed
face camber lines (Graph 6), it was decided to
discard the Johnson 5-terms line, because of its
strong thickness reduction of the profile area close to
the trailing edge, resulting in a rather poor sectional
modulus of inertia. The largest inertia modulus for
the profile is ensured by the Tulin-Burkhart camber
line, so this type of face line was the first to be
simulated. As in other conventional supercavitating
profiles, the trailing edge was sharply truncated and
with a trial and error procedure the thickness
distribution along the chord was calibrated in order
to obtain the maximum sectional area still remaining
in supercavitating conditions, as from Figure 6.

the pressure and velocity distributions around the


foil with the streamlines.

Figure 7 - Vapour Cavity distribution on first J3T design

Figure 8 - CP distribution and streamlines on first J3T design

Figure 9 - Velocity distribution on the first J3T design

Figure 6 - Cavity shape prediction on the T-B profile

With respect to the design 2D-CL=0.42 the obtained


T-B (Tulin-Burkhart) profile presented in Figure 6
(which the vapor phase is represented in blue and the
liquid phase in red) is only CL=0.36 with a
CD=0.035 and an efficiency of L/D= 10.3, too low
with respect to the ideal one for our scope.
It was then decided to go on with the Johnson Three
Terms (J3T) camber line for the face, designed for
the same bi-dimensional lift coefficient.
The first designed J3T thickness distribution (v.1),
as shown in Figure 7, was not a valid solution since
the vapor cavity does not start from the leading edge
of the profile and it is only formed at the maximum
curvature point of the back. Consequently the lift
predicted values are in good agreement with the
design values. In fact RANSE simulations predict:
v1:
CL=0.405 ; CD=0.040 ; L/D=10.1
(non optimal)
So a closer agreement of the lift value with the
theoretical (design) one, but still with a poor
efficiency similar to the T-B, for sure due to the
partly wetted back surface, which increases the
friction resistance. Following Figure 8 and 9 show
IX HSMV Naples 25 - 27 May 2011

Figure 10 - Vapor cavity on J3T v.2 design

Figure 11 - CP and streamlines distribution on J3T v.2 design

Figure 12 - Velocity distribution on the J3T v.2 design

A new tentative design (v.2) was then performed as


shown in Figure 10, in which the back line was
modified to increase the camber starting from midchord in order to anticipate the cavity detachment
point. Moreover the thickness at the leading edge
was also increased to achieve a larger area in this
region. Also in this case, the results are not
satisfactory: the cavity actually detaches earlier on
the back, but still is not created at the nose.
Predicted hydrodynamic characteristics for the
solution v.2 are:
v2:
CL=0.378 ; CD=0.044 ; L/D=8.6.
Notice that not the CD increases and CL decreases
with respect to the previous version v1. This is
because of the larger profile thickness at the leading
edge not balanced by a pressure recovery at the
trailing edge of the profile (which is inside the
cavity). The increase of curvature from midchord
station was not beneficial.

resulting in a drastic reduction lift, simply explained


by the change of curvature the initial part of the
back.

Figure 13 - Vapor Cavity shape on J3T v.3 design

Figure 16 - Vapor Cavity shape on J3T v.4 design

Figure 14 - CP distribution and streamlines on J3T v.3 design

Figure 17 - CP distribution and streamlines on J3T v.4 design

Figure 15 - Velocity distribution on J3T v.3 design

Figure 18 - Velocity distribution on J3T v.4 design

The third design(v.3), shown in Figure 13, returned


to use the back shape of v.1 with a further decrease
of thickness on the back at the leading edge, in order
to cause the cavity detachment; but this was not
achieved, according RANSE calculations because
the thickness of the foil is still too high at the nose
and the local pressure does not go under the vapor
tension.
Estimated hydrodynamic characteristics:
v.3: CL=0.365 ; CD=0.045 ; L/D=8.1

No appreciable separation is noted on the back of


this profile, as from Figure 18, at the contrary of
what can be noted for the other versions v.2 and v.3.
in Figure 12 and 15 respectively.
Diagrams in Graph 7 shows the pressure distribution
on the face and the back of v.1 in red, v.3 in green
and v.4 in blue. The main difference is of course in
the back distribution where the only one showing
constant - CP = 0 is the v.4, while the other two
tentative designs create a pressure recovery area on

IX HSMV Naples 25 - 27 May 2011

The fourth design tentative (v.4), shown in Figure


16, the back line was brought to the limit as regards
local strength. Indeed with this line a thin cavity is
obtained that spans the entire back of the profile, but
the thickness of the profile in the first half chord is
very low. Global results results:
V4.: CL=0.432 ; CD=0.035 ; L/D = 12.2,
In line with what expected from the idealized theory.
As can be noted from Figure 17, the long concave
region on the back of the profile triggers the cavity
separation and maintains it along the back up to
midchord, slowly increasing the local thickness up
to the midchord point at which the back changes
curvature and bends toward the face more gently
than the face does.

10

the back, which in the case of v.3 is able to influence


also the face pressure distribution close to the nose.

The design study presented is only a first attempt to


find a base section fort the 3D super-cavitating
hydrofoil and will be possibly improved, especially
for what concerns the strength issues, in occasion of
the next foreseen design exercises. Next activities o

AKNOWLEDGEMENTS

The present work has been performed in the


framework of the research grant N62909-10-1-7116
of the Office of Naval Research Global, dedicated to
the Hydrodynamic Design and Assessment by CFD
Methods of Hybrid SWATH/Hydrofoil Hulls for a
Super High Speed USV".
Graph 7 - CP distributions on the different J3T foil designs

8
6

CONCLUSIONS

Calibration and validation of a state of the art finite


volume RANSE solver based on the volume of fluid
method for the solution of the turbulent viscous flow
with vapor/water mixture and a simplified RayleighPlesset model to control the phase changes and
fraction distribution in each cell have been presented
and discussed in the paper.
The best accuracy obtained from the RANSE
simulation in terms of global force components is in
the order of 10% against the experimental values,
which is not particularly satisfactory, but still good
for engineering purposes and also taking also into
account the uncertainties levels of the experimental
results.
Further investigations need to be done including the
effect of surface tension (Weber number) and a even
further mesh refinement level. Some newer
reference experimental results would be also deemed
in order to avoid uncertainties due blockage and
symmetry effect which were not corrected in the
available experimental results.
Finally the paper also presents and offers the results
achieved by an hybrid method to effectively design
2D super-cavitating profiles at small finite cavitation
numbers (such as 0=0.05 in our case). The
approach is based on mixed use of the Johnson
asymptotic linear theory to design the face line and a
DOE (Design Of Experience) approach for the back
shape using systematic RANSE flow simulations.
The application of this new design approach leads to
very satisfactory results in terms of the lift and
hydrodynamic efficiency (L/D) predicted by the
RANSE calibrated model with respect to the best
achievable ideal performance.

IX HSMV Naples 25 - 27 May 2011

REFERENCES

Baker E. S. (1975), Notes of the Design of Two Supercavitating Hydrofoils, DTNSRDC Report SPD-479-13.
Brennen C. E. (1995), Cavitation and Bubbles
Dynamics, Oxford University Press, ISBN 0-19-5094093.
Brizzolara S. (2004). Parametric Optimization of SWATHull Forms by a Viscous-Inviscid Free Surface Method
Driven by a Differential Evolution Algorithm, 25th
Symposium on Naval Hydrodynamics, St. Johns,
Newfoundland and Labrador, CA, Vol V, pp. 47-64.
Brizzolara S., Bruzzone D. (2007), Hydrodynamic
Assessment and Optimization of New Fast Foil Assisted
SWAMH, Practical Design of Ships and Offshore
Structures, PRADS 2007. Houston., vol. 1, pp. 205-211.
ISBN/ISSN: 0-943870-04-6.
Brizzolara S., Curtin T., Bovio M., Vernengo G.,
(2011), Concept Design and Hydrodynamic Optimization
of an Innovative SWAT-Hull Form by CFD Methods,
NURC Internal Report NR-01-11-1.2.
CD-Adapco (2009), Star-CCM+ User and Theory
Manual, version 4.04.011
Johnson, V. E. (1957), Theoretical and Experimental
Investigation of Arbitrary Aspect Ratio, Super-cavitating
Hydrofoils Operating Near the Free Surface, NACA RM
L57I16.
Sauer, J. (2000), Instationaer kavitierende Stroemungen
- Ein neues Modell, basierend auf Fron Capturing VOF
und Blasendynamik, Dissertation, Universitaet Karlsruhe.
Tulin, M. P., and Burkart, M. P. (1955) Linearized
Theory for Flows About Lifting Foils at Zero Cavitation
Number. DTMB Rep. C-638.
Waid, R.L. & Lindberg, Z.M. (1957) Experimental and
theoretical investigation on a Super-cavitating Hydrofoil,
California Institute of Technology, Pasadena.

11

APPENDIX RESULTS OF VALIDATION ON FOIL TESTED BY WAID & LINDBERG (1957)

Figure 19 Pressure coefficients around the 2 terms


supercavitating foil (Waid and Lindberg , 1957) predicted by
the RANSE model at different cav. numbers and =6 degree.

IX HSMV Naples 25 - 27 May 2011

Figure 20 Velocity magnitude around the supercavitating


foil, predicted by the RANSE model at different cav. numbers
and =6 degree (inlet flow speed of 9.14 m/s).

12

Figure 21 Predicted cavity shapes on the 2 terms supercavitating foil of Waid and Lindberg (1957) used in the validation study
of the RANSE model.

IX HSMV Naples 25 - 27 May 2011

13